Letter to the editor: Confidence limits based on the first occurrence of an event by V. T. George and R. C. Elston, Statistics in Medicine, 12, 685–690 (1993)

✍ Scribed by Kai Fun Yu

• Publisher: John Wiley and Sons
• Year: 1998
• Tongue: English
• Weight: 88 KB
• Volume: 17
• Category: Article
• ISSN: 0277-6715
• DOI: 10.1002/(sici)1097-0258(19980430)17:8<945::aid-sim835>3.0.co;2-p

✦ Synopsis

The controversy over the biasedness of the estimator 1/n for prevalence p would have been happily resolved by the authors' reply to Liu's letter to the Editor in the 30 March 1997 issue of Statistics in Medicine, except for the very last statement ' 2 there does not exist an unbiased estimator of p 2 ', on page 712, which is unfortunately false.

Consider the statistic º which is defined as "1 if n is 1, and "0 if n is greater than 1. It is easy to see that º is an unbiased estimator of the prevalence p. This is another example that it may be foolhardy to prefer an unbiased estimator because other than being unbiased º does not seem to be better in any other reasonable sense than 1/n. For another example, see Efron.